THE PROFIT LAB // 4 Strategies to Optimize Assortment Planning
WEEK 3
I was recently working with a major retailer who expressed that they had so many forecasts available to them that it was hard to know which one to use. There is a forecast for marketing, for the catalog, for the website, one for the replenishment of goods at a low level, one for financial merchandise planning at a high level of merchandise, one for the distribution center, and the list continued. Which one do we use for planning? It was almost enough for them to throw their hands up and just base their plan on last year. I laughed and said that if they did that they would be on par with almost every other retailer out there.
Sadly, my experience has shown that to be true. While forecasting has, to a certain extent permeated the realm of higher level merchandise financial planning it has yet to make a real beachhead in assortment planning. I would argue that there is a lot of opportunity to be gained if the forecast is incorporated into the assortment planning process for determining store assortment breadth, depth, and whether or not items will be carried at all.
Assortment Review
Finding the balance between the benefit of utilizing a forecast in assortment planning or not partly depends on what you are forecasting. When the assortment plan is synonymous with an assortment review process or category review the benefits definitely align with utilizing a forecast. An Assortment review process is most typically used in long life items. Whether they be hardlines merchandise or long life softlines merchandise, such as jeans, the forecast can predict performance of an item with a high degree of accuracy. Traditional forecasting systems require a great deal of history to provide a forecast that has a confidence level that is high enough to be worthwhile to incorporate into the process. Items that have long life, often referred to as replenished items, typically have a confidence level that is high enough. So, the results of a demand forecast, which is a forecast that incorporates lost sales and available inventory, can be utilized by the planner to determine which items should be kept, which items should be deleted, and which items should be added or removed from a specific cluster. Typically this process is completed using only historical performance. However, trends that may not be perceptible when looking at historical performance can be seen in a forecast.
Determining the breadth of the assortment to a specific store, or cluster of stores can also be enhanced by forecasting. By using a forecast to best match a product with clusters that are most likely to sell the item profitably, it is possible to reduce overstocks and prevent markdowns.
Forecasting for fashion
Nobody would tell you that it’s easy to forecast for fashion or any short shelf life product such as cell phones or DVDs. Why is it so difficult to forecast fashion? There are a number of reasons, but the primary issue is short life. Traditional forecasting systems need long periods of historical activity to identify selling trends and begin producing results they have confidence in. Add the complexity of sized merchandise and the data is much too granular to draw SKU / store level conclusions from. Many have come up with complex algorithms, constraints and rules that attempt to address this issue. So retailers have adopted an alternative approach: consolidation. By consolidating the histories of many products that have similarities to the current product, we feel confident that the current product will behave as its predecessors have. For example, when allocating a new product to stores, it’s common to use a base data set of the product’s class, or alternatively, choose a “like item”. This of course is simply a surrogate to address the limitations of forecasting and store replenishment. Since the products don’t live long, we supplement our need for more historical selling time by applying our knowledge of similar products or product groups to give us more data. This allows us to begin seeing selling patterns. We then apply calculations that interpret the relationships in this base of data to derive a calculated recommendation.
These calculations are simpler than forecasting routines, but together with the additional merchandise that makes up the base of data, they are much less volatile and therefore return reasonably stable results. We review this result and change it based on other dimensions of data we analyze, assumptions and intuition. Having said that, there are forecasting systems that have been able to aggregate similarities in products, such as attributes, price points, or fashionability to give a semblance of accuracy to a forecast.
Tracking Life Cycles
Recently, a few companies have had success applying forecasting to fashion allocation. They have done this by combining advancements in technology with innovation in retail science to understand the relationships of behavior across many different products, store types, and levels. Two of these relationships that have shown some promise are lifecycle and strategies. Tracking the lifecycle of an item at a store level to see how that store behaves with a new product that has a short life has shown to be an excellent indicator of future item behavior. A typical product introduction has a curve to it over time that shows how quickly a new product takes off and how long it produces positive results. Mapping that behavior by store to new items gives a solid indication of how a similar new item will perform in the same location.
Product Strategies
With the knowledge of life cycles, product strategies and price points will give the forecast lots of historical data points. Another helpful tactic is to create product strategies. An item’s strategy is defined by how the product is expected to behave or by assessing why the item is in the assortment. Traffic drivers, loss leaders, fringe items and core items are all terms that are typically used to describe an item’s strategies. The combination of strategies and lifecycles starts to give us a preview of an item’s behavior by store once it is introduced. These can be used to help a planner determine where certain items will perform well in order to determine which clusters are best to receive the item.
Technology to simplify the complexity
With automated inventory management systems, the complex execution can be simplified. Since these systems also understand what you as an allocator are trying to achieve, they can execute to that automatically. Only when they cannot do what you’ve asked of them does the allocator need to intervene. Even then, issues are addressed using business logic rather than trying to manage complicated calculations, statistics or controls.
The same process can be applied to any new item, whether short life or long. By using a culmination of information similar to that product, a new product can be forecasted with enough accuracy that a planner can have a good recommendation as to where that product should be carried. For example, by knowing how fashion-forward an item is, the item’s color, price point, and attributes, such as sleeve length, the forecast can use a consolidation of similar items to forecast how that item will perform in a given store based on that store’s historical performance metrics. If we spend more time finding the data that most closely reflects the trending, lifecycle, seasonality and historical demand of the item we’re allocating, results ultimately improve. Once these metrics are known, a planner can determine if the item will positively impact sales or profit enough to carry it in the store.
Forecasting for localization
The benefits to localization are rarely disputed. All retailers to a matter of degree are attempting to place the optimal assortment in each store based on that store’s propensity to sell. By looking at history alone for a given store the localization process is simply not going to be optimized. In an earlier installment to this topic I wrote about the need for clusters to continually adjust to the behavior of the stores. Stores should not be locked into a particular cluster for an entire season/year but should shift as plans become actuals. Additionally, SKU rationalization or optimization, depending on your definition, needs to be a part of the localization process. As stores behaviors change, items need to be added or removed from the assortment in order to optimize the stores performance.
Forecasting should also be part of the localization process, although not as blatantly as dynamic clustering or SKU Rationalization. Rationalizing of the SKUs should be based, in part, on the forecast of the SKU / store rather than solely based on history. A stores assignment to a cluster should also utilize a forecast to cluster the stores given their expected behavior in the near term. As a caveat, this only works if you are re-clustering the stores on a weekly or monthly basis. Any further out than that and I would not trust the forecast’s accuracy.
Forecasting for depth
The hard part in using forecasting is attempting to determine whether or not to add an item to the assortment and deciding what stores the item will be ranged to. The much easier portion of the assorting process is in determining how many of the items to hold in the store in order to capture expected demand. A forecast can help determine the depth of the assortment and arguably have a greater impact to the performance of that assortment than helping to determine the breadth. By clustering stores together based on a forecast, the stores that are likely to perform similarly are going to be grouped. Presentation quantity is, of course, a consideration of the depth of the assortment. Typically the planner has the ability to determine how much product goes into the store and does so by the store volume cluster. Using the reliable wedge, the planner will typically put more in the larger volume stores than the smaller ones. However, if the forecast becomes more reliable, the amount of product that initially goes to the store can be refined to a more granular level so as to avoid over or understocks early in the product’s lifecycle. A good allocation or replenishment should be able to take care of it from there.
In Summary
It’s easy to argue that the forecasts at the SKU/Store level are too inaccurate to be of any use to the assortment planning process, but with some new thinking of how to forecast, significant value can be gained.
<< Previous post in series | Next post in series >>
Learn more
To learn more about assortment planning, be sure to check back weekly, or sign up below to receive email notices when this blog is published.
Subscribe to receive weekly updates of this series HERE»
Download this blog as a PDF»
For resources on assortment planning, visit: http://quantumretail.com/solutions/assortment-range-planning/resources/
4 people like this post.
Assortment planning is one of the first areas retailers should assess in order to increase profit and margin. I will be taking you through the top four strategies to optimize assortment planning including: SKU rationalization, clustering, forecasting and financial plans.
